A Thesis Submitted to the College of. Graduate Studies and Research. In Partial Fulfillment of the Requirements. For the Degree of Master of Science

Transcription

1 SPECTROSCOPIC ANALYSIS OF SELECTED SILICON CERAMICS A Thesis Submitted to the College of Graduate Studies and Research In Partial Fulfillment of the Requirements For the Degree of Master of Science in the Department of Physics University of Saskatchewan Saskatoon By Sam Leitch Keywords: soft x-ray spectroscopy, synchrotron radiation, silicon nitride, silicon oxynitride, density of states, electronic structure Copyright Sam Leitch, May All rights reserved.

2 PERMISSION TO USE In presenting this thesis in partial fulfilment of the requirements for a Postgraduate degree from the University of Saskatchewan, I agree that the Libraries of this University may make it freely available for inspection. I further agree that permission for copying of this thesis in any manner, in whole or in part, for scholarly purposes may be granted by the professor or professors who supervised my thesis work or, in their absence, by the Head of the Department or the Dean of the College in which my thesis work was done. It is understood that any copying or publication or use of this thesis or parts thereof for financial gain shall not be allowed without my written permission. It is also understood that due recognition shall be given to me and to the University of Saskatchewan in any scholarly use which may be made of any material in my thesis. Requests for permission to copy or to make other use of material in this thesis in whole or part should be addressed to: Head of the Department of Physics and Engineering Physics University of Saskatchewan Saskatoon, Saskatchewan (S7N 5E2) II

3 ABSTRACT Silicon ceramics are popular in both commercial applications and material research. The purpose of this thesis is to present measurements and analysis of four different silicon ceramics: α, β and γ phases of silicon nitride and silicon oxynitride using soft x-ray spectroscopy, which analyses the electronic structure of materials by measuring the absorption and emission of x-ray radiation. Absorption and emission spectra of these materials are presented, many of which have not be previously documented. The results are compared to model spectra and together they provide information about the electronic structure of the material. Assignments of emission features to element, orbital, and site symmetry are performed for each material. Combinations of silicon and nitrogen emission spectra provide insight into the strained bonding structure of nitrogen. It is concluded that p-dπ interaction plays a role in the bonding arrangement of nitrogen and oxygen sites within these structures. The emission features of non-equivalent silicon sites within γ-si 3 N 4 are identified, which represents some of the first analysis of same element, non-equivalent sites in a material. Silicon absorption and emission spectra were plotted on the same energy scale to facilitate measurement of the band gap. Since previously measured band gaps are not well represented in literature, the measured band gaps were compared to values predicted using DFT calculations. The band gap values are in reasonable agreement to calculated values, but do not vary as widely as predicted.

4 ACKNOWLEDGEMENTS I would like to thank all the people that helped me through this process. I want to thank my parents and family both pushing me when I needed a push. I want to thank Cara for putting up with me when I was pushed too hard. I want to thank the entire beamteam group for their collaboration and camaraderie. I especially want to thank my supervisor, Alexander Moewes for giving me the opportunity to pursue this research. Without his support, none of this would be possible. Work at the Advanced Light Source at Lawrence Berkeley National Laboratory is supported by the US Department of Energy (Contract DE-AC03-76SF00098). Alexander Moewes is a Canada Research Chair. IV

5 TABLE OF CONTENTS PERMISSION TO USE... II ABSTRACT...III ACKNOWLEDGEMENTS...IV TABLE OF CONTENTS... V LIST OF FIGURES...VII LIST OF TABLES...IX LIST OF ABBREVIATIONS... X 1 INTRODUCTION METHODS/THEORY Soft X-ray Spectroscopy Calculations Broadening Factors ALS Beamline MATERIALS α-si 3 N 4 /β-si 3 N γ-si 3 N Si 2 N 2 O RESULTS/DISCUSSION Si L 2,3 Absorption N K Absorption O K Absorption Si L 2,3 Emission V

6 4.5 N K Emission O K Emission Non-equivalent Sites Band Gap CONCLUSION BIBLIOGRAPHY APPENDICES APPENDICES A Density Functional Theory B Wien2k Calculations C Absorption and Emission Spectra of Silicon and Silicon Ceramics VI

7 LIST OF FIGURES Figure 2.1: Comparison of progression of XAS spectrum in γ-si 3 N Figure 2.2: Comparison of different supercell atoms to XES spectrum in γ-si 3 N Figure 2.3: Schematic overview of undulator beamline 8.0 at ALS.[39] Figure 2.4: The schematic sketch of the soft x-ray fluorescence endstation [40] Figure 2.5: Three methods of measuring the L 2,3 absorption spectra of SiO Figure 3.1: Crystal structure of α-si 3 N 4 [43] Figure 3.2: Crystal structure of β-si 3 N 4 [44] Figure 3.3: Crystal structure of γ-si 3 N 4 [21] Figure 3.4: Crystal structure of Si 2 N 2 O[51] Figure 4.1: Measured Si L 2,3 x-ray absorption spectra with corresponding calculated spectra Figure 4.2: Measured N K x-ray absorption spectra with corresponding calculated spectra Figure 4.3: Measured O K x-ray absorption spectra with corresponding calculated spectra Figure 4.4: Total electron yield O K absorption spectra for Si 2 N 2 O and SiO Figure 4.5: Measured Si L 2,3 x-ray emission spectra with corresponding calculated spectra Figure 4.6: Measured Si L 2,3 emission spectrum with Si orbital symmetry breakdown for α-si 3 N 4 LPDOS below the Fermi level Figure 4.7: Si L 2,3 emission spectrum and orbital/site symmetry LPDOS for γ-si 3 N 4 Si s and d states below the Fermi level Figure 4.8: Measured N K x-ray emission spectra with corresponding calculated spectra Figure 4.9: N K emission spectrum of β-si 3 N 4 with LPDOS breakdown for N states below the Fermi level Figure 4.10: Measured x-ray emission spectra with orbital symmetry breakdown of LPDOS of Si and N in β-si 3 N VII

8 Figure 4.11: Measured O K x-ray emission spectra with corresponding calculated spectra Figure 4.12: Measured Si L 2,3 absorption with calculated, site-selective XANES Figure 4.13: γ-si 3 N 4 Si L 2,3 emission measured for energies indicated in Figure Figure 4.14: Difference spectra for comparison of emission from non-equivalent sites in γ-si 3 N 4.[57] Figure 4.15: Band gap measurements using combined Si L 2,3 XAS/XES spectra near the Fermi level Figure 4.16: Excitonic peak removal in γ-si 3 N Figure 4.17: Relative energy shifting over the Si L 2,3 emission energy range Figure B1: Partitioning of unit cell into atomic spheres (I) and interstitial region (II) Figure B2: Comparison of N p symmetric LPDOS using two different calculations Figure B3: Comparison of Si s+d symmetric LPDOS using two different calculations.. 69 Figure C1: Si L 2,3 x-ray emission spectra of nine selected silicon materials Figure C2: Si L 2,3 x-ray absorption spectra of nine selected silicon materials VIII

9 LIST OF TABLES Table 3.1: Calculated structural properties and band gaps for the selected materials Table 4.1: Calculated and measured Si L 2,3 band gaps compared for each material IX

10 LIST OF ABBREVIATIONS LPDOS SXS XAS XES CB VB RIXS ARPES DFT OLCAO LAPW+LO XANES TEY TFY PFY MCD CVD MNOS Si tet Si oct localized partial density of states soft x-ray spectroscopy x-ray absorption spectroscopy x-ray emission spectroscopy conduction band valence band resonant inelastic x-ray spectroscopy angular resolved photo-emission spectroscopy density functional theory orthogonal linear combination of atomic orbitals linear augmented plane wave with local orbitals x-ray absorption near-edge spectrum total elector yield measurement total fluorescent yield measurement partial fluorescent yield measurement multi-channel detector chemical vapour deposition metal-nitride-oxide-semiconductor tetrahedral coordinated silicon octahedral coordinated silicon X

11 1 INTRODUCTION Ceramics are quite common in material research and development. They are formed by taking simple elements or molecular compounds and subjecting them to intense heat and pressure. By adjusting the ratio of the components as well as the heat and pressure applied, different materials can be produced with unique electrical and mechanical properties. The ability to tune the electrical and mechanical properties of a material makes ceramics useful for a wide range of applications. They can be found in mechanical applications such as cutting blades, packaging, and structural components, as well as electrical applications such as integrated circuits, lasers, and other optoelectronics. The most well know material used in integrated circuits is silicon. Pure silicon is a semiconductor that when doped and configured in the proper geometry, produces a wide variety of integrated electronic devices. Although pure silicon not strictly a silicon ceramic, it serves as a basis for the other materials discussed in this thesis. Silicon and doped silicon have been thoroughly discussed in literature for many years, including both theoretical calculations of the electronic properties such as optical properties and electronic structure done by Joseph Mullaney,[1] and experimental analysis such as band mapping done by Luning et al.[2] Recent literature has focussed on using nanoscale properties of silicon. By confining silicon to quantum sized nanostructures, the optical and electronic properties can be further tuned to meet the needs of designers.[3-5] Because silicon is widely used in the electronics industry, researchers have also attempted to develop ceramics that use silicon as a base. One of the widely studied silicon ceramics is silicon nitride. In the late 50s, studies of silicon nitride revealed that it could 1

12 exist in two different crystal structures.[6-8] The two materials were designated α-si 3 N 4 and β-si 3 N 4 based on the energy required to form them. At the time, it was determined that these two phases were functionally identical. However, further analysis with more accurate equipment revealed distinctly different mechanical and electronic properties, and they have been compared and contrasted ever since.[9-20] More recently, a third crystalline phase of Si 3 N 4 was discovered, designated γ-si 3 N 4.[21] This third phase has a cubic spinel structure, where silicon atoms take up both tetrahedral and octahedral cation sites. The octahedrally coordinated Si in this structure is unique to nitrogen ceramics and serves to increase the hardness of the material as well as to reduce the band gap. There are also theoretical studies into what other materials could be created with this spinel structure.[22, 23] Although there are a large number of publications on the physical properties on γ-si 3 N 4,[24-27] studies of the electronic properties of this material are relatively scarce.[28, 29] Silicon oxynitride is another material that has been under investigation in literature. Si 2 N 2 O can be considered an intermediate crystal structure between Si 3 N 4 and SiO 2, only varying from the crystal structure of α-si 3 N 4 to account for the inclusion of O atoms. Si 2 N 2 O is considered a promising candidate for electro-optics and organic light-emitting devices.[30, 31] Since the benefits of Si 2 N 2 O have not been realized until recently, there are very few publications on the optical and electronic properties of this material as well.[32] The purpose of this thesis is to compare and contrast results of x-ray spectroscopic data taken for these four silicon ceramics: α-si 3 N 4, β-si 3 N 4, γ-si 3 N 4 and Si 2 N 2 O. X-ray 2

13 absorption and emission spectra for all transition energies within the range of beamline at the Advanced Light Source are presented, many of which have not been previously published. Comparison of the x-ray absorption and emission spectra with calculated spectra will be used to probe the electronic structure of each material based on element, orbital symmetry, and local geometry. The bonding structure of the materials will be analyzed as it relates to element, site and orbital symmetries. Finally, by combining the absorption and emission on the same energy scale, a band gap measurement for each material will be determined. Since determining the band gap using x-ray emission and absorption is not widely used, the advantages and disadvantages will be discussed which provide uncertainty in the measurement. Since measured band gaps of these materials are not well represented in literature, the band gaps will be compared to calculated band gaps obtained using density functional theory. 3

14 2 METHODS/THEORY 2.1 Soft X-ray Spectroscopy Soft x-ray spectroscopy utilizes electron excitation to analyze materials.[33] Photons in the soft x-ray energy region (50 ev 10 kev) excite electrons from low-lying core states to unoccupied conduction band (CB) states. The absorption and subsequent emission of photons produce distinctive energy dispersive spectra. Soft x-ray emission and absorption spectroscopy have many useful properties that make it a powerful tool in the field of materials research. SXS is element specific, since different elements have different core-cb transition energies that rarely overlap. It is bulksensitive, since soft x-rays have a penetration depth of nm well beyond the range of surface-sensitive states. Core orbitals are highly localized so it provides only information surrounding individual atomic sites. Lastly, only states that follow a specific orbital symmetry will be measured since the process is dipole in nature, following selection rule l=±1 where l is the orbital quantum number. When discussing soft x-ray spectra, we will quite often refer to the LPDOS or localized partial density of states. SXS deals with exciting electrons so the spectra produced are related to the density of electrons or electronic states in a given energy range. Since SXS is site specific, we are only interested in the electronic states localized around a single atomic site. SXS is also sensitive to atomic orbital symmetry so we are only interested in part of the total density of states surrounding the atom. SXS is therefore related to the localized partial density of states (LPDOS) surrounding a single atomic site within a given energy range. 4

15 X-ray absorption spectra (XAS) are produced when an electron is excited from a core hole to higher lying orbitals or, as in the case of light elements like Si and N, to the conduction band. Since core electrons are localized to their atomic sites and have atomiclike structure with well-known binding energy, the probability of absorption closely resembles the LPDOS of unoccupied states in the conduction band. By varying the excitation energy and measuring the rate of absorption, an XAS spectrum is obtained. When an electron is excited to conduction band states, a core hole is left in the atom. The hole is filled with an electron from occupied valence band (VB) states. Among other nonradiative processes, the energy from this transition can be released by emitting a photon. Since low-lying core states have well-known energies, the emission spectrum resembles the LPDOS of occupied states below the Fermi level. This process is known as x-ray emission spectroscopy (XES). X-ray emission is an improbable process, and requires long count times or high incident beam intensity to provide the required measurement statistics. For example, Si L 2,3 emission for Si with one 2p hole will occur with a probability x 10-4, while N K emission has a probability of 5.2 x 10-3.[34] Resonant emission is used to combat this. By tuning the excitation energy to features in the XAS spectrum a resonant absorption/emission process is created, which produces a large number of core holes and therefore an increased emission intensity. There is a separation commonly followed when discussing XES spectra that revolves around whether absorption and emission can be considered a single process or a combination of processes produced by independent states.[35] When absorption and 5

16 emission are considered separate processes, the system is allowed to remain in a pseudostable excited state for a finite amount of time. This is case for non-resonant excitation when the excitation energy is sufficiently above the ionization threshold. During that time, random phonon interactions alter the momentum of the core hole and the resulting emission process is considered to have random momentum. When we discuss XES spectra, we are generally referring to spectra that are produced through this two-step process. Alternatively, resonant inelastic x-ray spectroscopy (RIXS) is the study of materials using combined excitation/emission process. The material absorbs a photon, and a second photon is emitted at a related energy without the appearance of a pseudo-stable intermediate state. Since photons transfer very little momentum to the system, the momentum during the combined absorption/emission process remains constant. RIXS measurement and analysis provides a greater degree of control for measurements, since both momentum and energy difference are known. The excitation energy of the system influences the emission spectrum of the material. A further degree of control can be gained if the sample is a single crystal with known geometry. Angular resolved photoemission spectroscopy (ARPES) is a technique that uses photons to excite electrons out of a material into the continuum. When a single crystal sample is used, the momentum of emitted electrons is exactly known. Such angular resolved techniques can be used, in conjunction with RIXS, to map the complete band structure of materials. Unfortunately, new materials often require high temperature and pressure to synthesize and single crystal samples are rare. 6

17 2.2 Calculations Our group has relied on collaborators to perform the complex density functional calculations required for theoretical comparison. In recent literature, one of the most published authors on the subject of simple structured ceramics (especially silicon ceramics) is Prof. Wai-Yim Ching at the University of Missouri - Kansas City. He has generously provided our group with calculated data that has proven profoundly informative and has become the basis for much of our data analysis. Even though these calculations are not performed by our group it is important to have a strong understanding of what calculations can be performed and, most importantly, what limitations they have. Knowledge of the advantages and limitations of calculations allows comparisons of calculated and experimental results. Much of the recent work in modelling solid-state systems has revolved around the use of density functional theory (DFT). For a more detailed analysis of DFT calculations, please refer to Appendix A. This calculation method involves taking the time averaged electron density for occupied states in the system, and minimizing the overall energy functional. This is generally accomplished by solving for a linear set of electron Schrödinger equations for electron-electron and electron-nuclear Coulomb potentials. To accomplish this, Prof. Ching s group uses the orthogonal linear combination of atomic orbitals method (OLCAO),[36] which uses a Bloch sum of atomic or atomic-like orbitals centred on atomic sites. By minimizing the energy functional of the orthogonal equations, the ground state energy eigenvalues and wave functions can be determined. From the wave 7

18 functions and corresponding eigenvalues, properties such as band structure and LPDOS can be derived. Since Bloch sums are used in the analysis of the system, band structure is the most fundamental property obtained. The band-structure can then be further broken down, based on atomic, orbital and spin symmetries. LPDOS can then be derived based on the same criteria. An example of the unoccupied LPDOS of γ-si 3 N 4 with nitrogen p orbital symmetry is shown in Figure 2.1(a). (d) Experimental XAS Intensity (arb. units) (c) XANES (b) Corehole LPDOS Intensity (arb. units) (c) Experimental XES (b) Ground State LPDOS (a) LPDOS (a) Core Hole LPDOS Energy (ev) Emission Energy (ev) Figure 2.1: Comparison of progression of XAS spectrum in γ- Si 3 N 4 Figure 2.2: Comparison of different supercell atoms to XES spectrum in γ-si 3 N 4 LPDOS gives a basic trend for the experimental data, but lacks the effects of the core hole on the wave functions. The core hole left after an electron is excited to conduction band states will interact with and change the LPDOS surrounding that nucleus. To calculate the effects of core hole interaction, a supercell approach is used. Instead of 8

19 calculating wave functions for a single unit cell, a number of unit cells are combined, limited to ~100 atoms by the available computing power. An electron is removed from the core of a single atom in the supercell, and placed in the lowest unoccupied states. The wave functions of the new system are calculated, and the resulting LPDOS are derived. Figure 2.1(b) shows an example supercell LPDOS for an N atom in γ-si 3 N 4 with a 1s core hole. The core hole included LPDOS more accurately represents the experimental absorption spectra, as it more accurately represents the availability of excitation states. However, this picture does not take into account variations in transition probability. The probability of exciting an electron from a core state to a conduction band state at a given energy is known as the transition probability. Using the same supercell approach, wave functions for both the initial and final states of the excited system can be performed. By including transition matrix elements, a complete quantum-mechanical interpretation of the x-ray absorption near-edge spectrum (XANES) is derived. Figure 2.1(c) shows the XANES for the same N K absorption of γ-si 3 N 4. This data can be directly compared to experimental data (Figure 2.1(d)), as it includes all major components of a quantum-mechanical interpretation of an x-ray absorption spectrum. The final step in the process would be to model emission spectra based on excitation energy. Prof Ching s group does not currently do this type of calculation, but is in the works. In the meantime, LPDOS data is used for comparison. It would seem most logical that the LPDOS that includes core hole interactions would be the closest representation of the emission spectrum, since it represents the wave function 9

20 of the atom with a core hole. Figure 2.2 shows (a) the LPDOS of N in γ-si 3 N 4 for the excited atom in a supercell, (b) The LPDOS of an atom that remains in the ground state in the same supercell, and (c) the experimental emission spectrum for comparison. The LPDOS of the two ground state and core hole N atoms are dramatically different and the ground states LPDOS appears to more accurately model the measured emission spectrum. This is counter-intuitive at first, and can be explained the same way as for absorption: the emission spectrum represents the final state of the system. In this case, the final state of the system is the ground state. All absorption/emission processes tend to exhibit properties of the final state of the system, as explained by Vonbarth et al.[37] This effect is known as the final state rule. 2.3 Broadening Factors As with any kind of spectroscopic study, broadening factors in XES/XAS limit the accuracy and detail of analysis. However, broadening can also be used as a tool to learn new properties of a material that may otherwise be unobtainable. There are three broadening effects we take into account when modelling XES/XAS: Instrumental broadening, core hole lifetime broadening, and final state lifetime broadening. While these are not the only broadening effects involved in XAS/XES, these three factors have been isolated as the major and measurable causes of spectral broadening in these systems. Instrumental broadening is a statistical process caused by the optics and the environment that measurements take place. Purity of sources, purity of optical components and duration and intensity of measurements all contribute to instrumental broadening. Instead of discussing in detail each effect and how much it contributes, they are generally lumped 10

21 into a single broadening factor called resolving power (E/ E). Instrumental broadening is gaussian in nature and is modeled in calculated spectra by a gaussian function with a standard deviation derived from the resolving power. Core hole lifetime broadening is caused by the short lifetime of the core hole created in the photon absorption process. The short lifetime causes a broadening in the energy of the state based on the Heisenberg uncertainty principle. Core hole lifetime broadening is modeled using a lorentzian function with a constant broadening factor based on the documented lifetime of the core hole.[38] Final state lifetime broadening is very similar to core hole lifetime broadening. It is a result of the limited lifetime of the system after absorption or emission. After absorption, there is an electron left in the conduction band. After emission, there is a hole left in the valence band. Each of these excited states has a limited lifetime before they decay by either radiative or non-radiative processes. This lifetime causes an uncertainty in that state s energy value, broadening its respective spectrum. Because different energy states have different average lifetimes, final state broadening is not constant over the spectral range. Final state lifetime broadening is also simulated with a lorentzian function with a broadening factor derived by Goodings and Harris[39] and given by: E W 1 E E 0 E F E 0 2 (2.1) Where W defines the scale of the broadening, E F is the Fermi energy, and E 0 is the energy at the bottom of the lowest conduction band. All our calculated XES spectra use this equation for final states broadening, with a broadening scale (W) of 1 for both Si and N. 11

22 Emission spectra used in this thesis have been broadened according to these three factors. This was accomplished by taking the combined valence band LPDOS for each material based on the element and orbital symmetries measured and applying both a gaussian and lorentzian broadening routine I developed for Origin using the previously discussed values as inputs. Calculated XANES spectra used in this discussion have been previously broadened by a constant gaussian function, which is sufficient for the purpose of comparison. 12

23 2.4 ALS Beamline Figure 2.3: Schematic overview of undulator beamline 8.0 at ALS.[40] Figure 2.4: The schematic sketch of the soft x-ray fluorescence endstation [41] All measurements to date have been performed at Beamline of the Advanced Light Source in the Lawrence Berkeley National Laboratory[42]. Beamline 8.0 (Figure 2.3) is an undulator beamline, capable of producing photon energies of ev, with an intensity of photons per second at a resolving power (E/ E) of Radiation from the undulator is monochromatized using a grazing incidence, spherical grating monochromator. The water-cooled exit slit is moved so that it and the entrance slit satisfy 13

24 Rowland geometry in order to provide optimal focus. A horizontal refocusing mirror is used to compress the beam and refocus the incident radiation on one of four possible endstations. Beamline endstation (Figure 2.4) is a bell jar-type sample chamber with a spectrometer mounted at a right angle to the incident beam. The spectrometer consists of an entrance slit, four interchangeable spherical gratings, and a multi-channel plate detector that are also positioned based on Rowland geometry. We use a 50 µm entrance slit located in the sample chamber approximately 1 cm from the sample. This spectrometer configuration allows energy-dispersive XES spectra to be taken, with a resolution of 150 mev at 100 ev, and 700 mev at 400 ev. In addition, the endstation can be configured to perform both XAS and XES measurements without removing the sample from vacuum. True absorption of a thin sample would provide the most natural representation of the XAS of the material. However, at 100 ev photon energy, Si 3 N 4 has an attenuation length of 5 nm.[43] Because absorption probability increases exponentially with sample thickness, a freestanding sample under observation would have to have a known, uniform thickness much less than 10 nm to expect consistent, quantifiable measurements. In addition to being extremely fragile, samples with such small thickness would provide mostly surface sensitive results. As alternatives, beamline provides three measurement options for XAS (Figure 2.5). 14

25 Figure 2.5: Three methods of measuring the L 2,3 absorption spectra of SiO 2. The total electron yield (TEY) method involves measuring the current required to neutralize the ionized sample as electrons are removed by the incoming photons. The core hole produced by absorption can be filled by an electron from the occupied states. The electron filling the core hole can give its energy to a neighbouring electron, which is kicked out of the material into the vacuum. The number of electrons being emitted is proportional to the number of electron falling into core holes, which is proportional to the number of core electrons being excited, which is proportional to photon absorption. However, electrons have a much stronger attenuation length than photons, so not all of the electrons can escape. TEY is therefore more surface sensitive than pure photonin/photon-out methods. This method is also problematic with insulating samples, which show charging affects that skew actual absorption. 15

26 Partial fluorescence yield (PFY) involves measuring the emitted radiation accepted by the spectrometer detector. This method therefore provides a measure of the absorption that produces emission within a given energy range. Emitted photons do not suffer from the surface sensitivity associated with TEY. In addition, this method is very useful if one is only interested in electrons emitting within a given energy range. However, since emission is not an efficient process, the spectrometer entrance slit is so small and we are only measuring a limited emission energy range this process requires long measurement times in order to get a measurement with a reasonable amount of noise. The PFY measurement displayed in Figure 2.5 was taken using the same time slices as the TEY and TFY spectra. The results are so noisy that only basic data can be extracted. Longer count time would have produced better results. Recently, a channeltron was installed in the sample chamber that provides total fluorescent yield (TFY) measurements. This device uses a cascading scintillation process to convert photons (or electrons) into electronic pulses. The head of the channeltron is biased to repel incident electrons, so only photons are measured. Emitted photons do not suffer from shorter attenuation length, which means they produce more bulk-sensitive spectra, and the channeltron measures a wide range of photon energies, increasing intensity and therefore measurement statistics. The emitted photons are still susceptible to self-absorption, which is both a blessing and a curse. Self-absorption causes spectral spreading and reduced intensity as the excitation energy is tuned across additional thresholds (to higher absorption energies). This has a negative impact over the entire energy range of absorption but in the same respect, since the self-absorption is minimal at the absorption onset, TFY increases the intensity and resolution at the absorption onset. 16

27 The absorption onset often contains the most interesting features within absorption so this feature is actually quite advantageous. 17

28 3 MATERIALS The four materials discussed in this thesis are α-si 3 N 4, β-si 3 N 4, γ-si 3 N 4, and Si 2 N 2 O. Some basic properties of the materials are listed in Table 3.1. Table 3.1: Calculated structural properties and band gaps for the selected materials Material α-si 3 N 4 [12, 44] β-si 3 N 4 [12, 44] γ-si 3 N 4 [23, 44] Si 2 N 2 O[12, 44] Space Group P3 1 /c (159) P6 3 /m (176) Fd 3 /m (227) Cmc2 1 (36) Lattice Constants (Å) a = c = a = c = a = a = b = c = Si coordination 4 4 4, 6 4 N coordination Average NN Length (4) (Å) (6) E Gap minimum (ev) E Gap direct at Γ (ev) (Si-N) (Si-O) 3.1 α-si 3 N 4 /β-si 3 N 4 Figure 3.1: Crystal structure of α- Si 3 N 4 [45] Figure 3.2: Crystal structure of β- Si 3 N 4 [46] 18

29 α-si 3 N 4 and β-si 3 N 4 are the lowest energy phases of silicon nitride, and until recently have been considered the only phases of Si 3 N 4. They are very similar in structure and function, so they are often lumped together when discussing nitrides in literature. α- Si 3 N 4 (Figure 3.1) is characterized by a trigonal lattice structure. It contains Si atoms that are each connected to four N atoms to form a tetrahedron, and N atoms surrounded by three Si atoms in a planar arrangement (trigonal). The entire structure is planar with stacking pattern ABCDABCD β-si 3 N 4 (Figure 3.2) is characterized by a hexagonal lattice structure. It also contains Si atoms that are surrounded by four N atoms in a tetrahedral coordination, and N atoms surrounded by three Si atoms in a trigonal arrangement. The overall structure is planar, with stacking pattern ABAB Even in large crystal grain sizes, electronic properties of α-si 3 N 4 and β-si 3 N 4 resemble amorphous Si 3 N 4. The resemblance of the materials is so good that there is a consensus that amorphous Si 3 N 4 naturally forms a mixture of polycrystalline Si 3 N 4 phases when pressed. This is most likely due to the strong localization of the silicon sites, which will be discussed further in Section 4.1. Mechanically, both of these materials are very hard and resistant to oxidation. These two factors make Si 3 N 4 ceramics useful for cutting tools and other industrial applications. Although each crystal itself is very hard, the overall ceramics containing α and β-si 3 N 4 are relatively brittle. This has been attributed to the crystal grain interface with the ceramics glassy region. Crystal growth research has overcome this problem by creating β-si 3 N 4 whiskers. [47] These elongated crystal grains along with an overall bimodal grain size make β-si 3 N 4 much stronger. However, producing the same structure with α- Si 3 N 4 has proven more difficult. 19

30 Crystal growth researchers have been experimenting with different ways of further increasing the hardness and resistance to intense heat of these materials. It has been found that the introduction of aluminium and oxygen into the ceramics can produce such results. These new class of ceramics are known as SiAlON ceramics. Much of the literature on crystal growth of silicon ceramics is currently focussed on such SiAlONs. Electronically, α-si 3 N 4 and β-si 3 N 4 are considered insulators. They have wide indirect band gaps (4.67 and 4.96 ev respectively [12, 44]) and are used as both insulators and protective layers in integrated circuits. The most common method of applying Si 3 N 4 to a silicon wafer in integrated circuits is chemical vapour deposition (CVD), which creates a generally amorphous thin film of material. When used as an insulator, nitrides can produce a quantum well used to trap electrons, or increase switching voltages. Such metal-nitride-oxide-semiconductor (MNOS) devices are common in non-volatile memory and certain solar cells.[48] When used as a protective layer, Si 3 N 4 can be found in FinFET applications [49] or used as a diffusion barrier. In this study, a polycrystalline sample of α-si 3 N 4 is used. It was measured in both pressed and powered forms, with identical results. The sample is yellow-grey in colour and translucent. For β-si 3 N 4 a single crystal sample was used. The sample was transparent, and measured approximately 2 mm x 2 mm x 0.1 mm. 20

31 3.2 γ-si 3 N 4 Figure 3.3: Crystal structure of γ-si 3 N 4 [21] Originally synthesized by Zerr et al[21] in 1999, γ-si 3 N 4 is the highest-energy phase of Si 3 N 4 requiring an environment of 13 GPa and 1800 K to form. This phase is harder than the previous two phases and, when it was first synthesized, was second only to diamond as the hardest known material. In addition to being very hard, this nitrogen ceramic is extremely resistant to oxidation. These two properties alone make γ-si 3 N 4 valuable for mechanical devices such as cutting blades. γ-si 3 N 4 has a face-centred cubic spinel structure (Figure 3.3) in which nitrogen atoms surround Si atoms in two different arrangements. Six nitrogen atoms surround octahedral silicon sites while four surround tetrahedral sites. The octahedrally bonded silicon is unique to γ-si 3 N 4. In fact, although observed frequently in oxide ceramics, γ-si 3 N 4 was the first nitride ceramic where this six-fold arrangement of N anions was observed. This discovery prompted a flurry of research into materials that could benefit from a spinel 21

32 structure, including spinel-phase C 3 N 4, which is predicted to be harder than diamond, but has proven to be difficult to synthesize in a pure form.[50] According to ab-initio calculations, γ-si 3 N 4 is a wide-gap semiconductor, with a direct band gap of 3.45 ev.[23] Ultraviolet emission from this band gap would be useful for light emitting devices, as it can either be used directly or changed to visible light using phosphors. Likewise, semiconductor-based power systems that are based on wide-gap semiconductors benefit from increased efficiency, reliability, and temperature tolerance. However, experimental verification of this band gap has not been published, which is a straightforward process using SXS. In this study, a polycrystalline sample of γ-si 3 N 4 is used. The γ-si 3 N 4 powder was synthesized by Toshimori Sekine from β-si 3 N 4 by the shock compression method[51] and purified by a hot hydrofluoric acid solution.[52] The powder has been characterized using x-ray diffraction, Differential Thermal Analysis[53] and 29 Si magic-angle-spinning NMR.[28] The oxygen content of the purified γ-si 3 N 4 has been determined to be 2.0 wt% by a high temperature combustion method in the LECO Co. furnace. Therefore, the γ- Si 3 N 4 powder is considered to consist of pure nitride spinel phase. The sample is known to have a grain size of nm. 22

33 3.3 Si 2 N 2 O Figure 3.4: Crystal structure of Si 2 N 2 O[54] Although Si 2 N 2 O has been around for many years, its uses have been limited until very recently. Si 2 N 2 O is characterized by an orthorhombic lattice structure. The overall structure is a network of sheets of Si 2 N 2 interconnected by O atoms. N atoms are still surrounded by three Si atoms in a trigonal structure while Si retains its tetrahedral bonds with one of the N atoms being replaced by O. The unit cell for Si 2 N 2 O is pictured in Figure 3.4. The reason Si 2 N 2 O has not been widely used until recently is its very wide band gap. At 5.20 ev, it is very much an insulator. It is not as hard as α-si 3 N 4 or β-si 3 N 4, so has limited mechanical applications. However, Si 2 N 2 O has been proposed for use in integrated optical communications devices.[31] By bombarding a quartz substrate with nitrogen, Si 2 N 2 O can be integrated into the material. By adjusting the amount of N introduced into the material, the index of refraction of the material can be tuned 23

34 creating a planar wave-guide. This technique has been used to produce polarization controllers, couplers, filters and switches. In the present study, a polycrystalline pressed and powdered sample of Si 2 N 2 O was used. It is white-grey in appearance and exhibits no glossy finish. 24

35 4 RESULTS/DISCUSSION 4.1 Si L 2,3 Absorption Figure 4.1: Measured Si L 2,3 x-ray absorption spectra with corresponding calculated spectra. The most common and most published spectra available for these materials are the Si L 2,3 spectra (both absorption and emission). α-si 3 N 4 and β-si 3 N 4 absorption spectra have been well published[14, 20, 29, 32, 55-59] while both γ-si 3 N 4 [29, 60] and Si 2 N 2 O [61] have only a few published results. Figure 4.1 shows our XAS spectra for the four materials, along with the associated calculated XANES spectrum. Vertical lines placed on the 25

36 spectra indicate areas of interest. Discounting the ratio of peaks, three of the experimental spectra: α-si 3 N 4, β-si 3 N 4, and Si 2 N 2 O have very similar spectral features in terms of energy. This emphasizes the importance of local coordination of Si in these types of materials, which is much stronger than the influence of alternate elements, as is the case with Si 2 N 2 O, or overall crystal structure changes, as is the case with α-si 3 N 4 /β-si 3 N 4. γ- Si 3 N 4 is the only material to show a substantial spectral deviation. It is also the only material to exhibit a change in the local coordination of Si atoms with the introduction of octahedral Si. By comparing the three similar experimental spectra to that of γ-si 3 N 4, it appears that the inclusion of octahedral Si causes a superposition of two spectra: Si tet and Si oct. This conclusion seems valid, again considering the highly localized nature of Si L 2,3 absorption spectra. Since the experimental spectra are similar, in order to discuss the features of each spectrum we need only discuss the general shape of all of the experimental spectra. The vertical line at ev indicates an area of the spectra due to the onset of the conduction band DOS. Sharp peaks characterize the onset, which are not found in the onset of crystalline Si.[56] These peaks have been attributed to localization of states in the conduction band. The localization effect increases the lifetime of electrons that are excited to these states, therefore reducing lifetime broadening and producing sharp peaks. These localization effects have been documented for SiO 2, as well as other ceramics not containing Si.[62] O Brien et al show that these states are indeed due to conduction band states by measuring their emission. Since these states are nearest the Fermi edge, they can be populated thermally. Since they are populated, there is a chance for emission to occur. It is interesting to note the similarities of the peak structure within this energy range for 26

37 each of the materials. All three materials exhibit a 4-peak structure, with similar peak ratios and relative energy positions. The only observable change is the overall location of the peaks, which is shifted to lower energy in β-si 3 N 4 and higher energy in Si 2 N 2 O compared to α-si 3 N 4. The peak resolvability also changes for each spectrum. This resolvability or sharpness is most likely due to manufacturing processes of the materials measured as well as measurement factors rather than effects of the materials themselves. After the localized onset, the spectra exhibit a delocalized band-like L 2,3 onset. This absorption feature is broad compared to localized onset discussed earlier, which is due to the short lifetime of electrons in these delocalized states. There is very little to resolve from this peak, other than it s location and relative intensity. When comparing to the calculated spectra, it should be noted that this area is not very well represented. This may be due to the inability of the calculations to account for the localization of the near-edge states or it may simply be washed out due to the constant broadening factor used in XANES. In a similar respect, the onset of L 1 absorption can be seen above 112 ev. This peak is very broad, and similar in nature to the L 2,3 band onset. When comparing the measured and calculated spectra, each of the calculated XANES show a peak that corresponds in energy to the L 1 onset, although it appears that the calculations over-estimate its influence on the absorption spectra. The Si L 2,3 absorption spectrum of γ-si 3 N 4 is different from the absorption spectra of the other materials, which is attributed to the presence of octahedrally coordinated Si. Si oct 27

38 produces a different absorption spectrum than Si tet. Since γ-si 3 N 4 contains both tetrahedral coordinated Si and octahedral coordinated Si, the absorption spectrum is a superposition of absorption from these two non-equivalent sites. The peak appearing under the vertical line at ev in the γ-si 3 N 4 absorption spectrum is the L 2,3 onset for Si oct. This peak overpowers at least two of the four peaks of the localized tetrahedral onset since there are twice as many octahedral Si sites as tetrahedral. Si oct does not appear to exhibit the same localization features found in Si tet. The two sharp onset peaks under the vertical line located at ev are also present only in γ-si 3 N 4. These peaks are highly localized, but they are not attributed to localization of the conduction band as with the Si tet onset. Instead, these peaks are attributed to newly created excitonic states within the band gap. Excitonic states are electron-hole pair states that are created by coulombic interaction. When an electron is excited from a core state, it is excited to a state localized around the atomic site. Since the electron and the newly created core hole both represent charged particles (or virtual charged particles), they exhibit coulombic interaction. This interaction can create electronic states that would otherwise not exist in the system, including states that exist within the band gap. 28

39 4.2 N K Absorption Figure 4.2: Measured N K x-ray absorption spectra with corresponding calculated spectra. The N K absorption spectra as well as calculated XANES are shown in Figure 4.2. Vertical lines placed on the spectra indicate areas of interest. A literature search revealed few previously documented N K spectra of silicon nitride, and only of amorphous samples created by bombarding Si with N atoms. [63, 64] These spectra therefore represent some of the first N K spectra of these materials. The gradual onset of the γ- Si 3 N 4 absorption spectrum at ~402 ev is lower than the other three materials. This early 29

40 onset causes a reduced band gap that is consistent with band gap values discussed in Section 4.7. The N K onset for the other three materials is much sharper, and comes to a peak located near ev. All of the spectra exhibit a strong peak under the vertical line located at ev except β-si 3 N 4, where it can be resolved into two separate peaks. This is corroborated with the calculated spectra, and is a result of the band structure in that energy range. It is interesting to note that this same splitting appears in the calculated XANES for γ-si 3 N 4, even though it does not appear in the measured spectrum. There is a shoulder located near 410 ev on all of the spectra but γ-si 3 N 4, where it is yet another peak. The location of this feature is reproduced in the calculated XANES, however the intensity is not. The extended absorption spectra exhibits additional small peaks near ev and ev. These features are remarkably well reproduced in the calculated spectra. Finally, there is a bump in each spectrum located above 420 ev, which coincides with the L 1 absorption edge for the Si spectrum. Unlike Si, N spectra differ significantly for different materials. However, the fact that the spectra differ for each material is not as interesting as the existence of observable finedetail superimposed on the onset. Nitrogen absorption spectra have been used in the past to determine nitrogen content of amorphous SiN x films.[63, 64] It turns out that an excitonic peak appears in the spectrum that increases as a function of nitrogen content. The important part to note about these studies is the shape of the nitrogen spectra. The amorphous SiN x spectra have one large onset with no observable variation, while the nitrogen spectra shown in Figure 4.2 have observable features within the same energy range. This means that electronic states surrounding the N atoms must be influenced by the crystal structure as a whole. 30

41 The fact that the N spectra have more variation for a crystal then for an amorphous sample is somewhat counter-intuitive. When atoms arrange in a crystal, the outer band electrons switch from localized atomic orbitals to delocalized band states. This delocalisation process increases the density of states possible in a given energy range, thereby decreasing the lifetime of electrons in any given state and increasing the broadening factor. The absorption spectra for a given element should become more washed out in a crystalline material then an amorphous material, which is not the case for the nitrogen K edge. Instead, the N K edge appears to be sensitive to the long-range order associated with crystal structures, which can be exploited to determine the overall structure of an unknown material. 4.3 O K Absorption Figure 4.3: Measured O K x-ray absorption spectra with corresponding calculated spectra. 31

42 The oxygen K edge absorption spectra are shown in Figure 4.3. Both TEY and TFY measurement methods are included, along with the calculated XANES spectrum. O K absorption spectra of Si 2 N 2 O have only recently been documented, and those spectra were performed on only amorphous thin film samples. These spectra therefore represent the first O K emission spectra of polycrystalline Si 2 N 2 O. The spectra exhibit three preedge features that appear between 531 ev and 536 ev, a sharp onset at ~536 ev, followed by a gradual peak at ~540 ev with limited features. The two measured spectra agree with the calculated XANES spectrum (bottom) excluding the pre-edge features. The short vertical line at ev represents the calculated onset of the conduction band, which was derived by aligning the O K emission spectrum to the calculated LPDOS spectrum and extending it beyond the Fermi energy to the onset of the conduction band. While this method does not take into account any spectral shifting due to the presence of the core hole, it gives a good approximation of the location of the calculated conduction band onset. It is clear that the pre-edge features exist entirely within the band gap; however, it is not clear whether the pre-edge features are due to impurities in the apparatus or actual material properties that are not modelled in the calculated XANES. In order to clarify this, we compare our spectra to those of α-sio 2 and amorphous Si 2 N 2 O thin films measured by McGuinness et al.[32] The O K absorption spectra for these two materials are similar to our acquired spectrum of poly-crystalline Si 2 N 2 O, excluding the appearance of the pre-edge features. This means that either the pre-edge features are impurity based, or crystalline Si 2 N 2 O exhibits properties that no other Si-O ceramic exhibits, the later of which is very unlikely. 32